pith. sign in

arxiv: 2606.06154 · v1 · pith:HDKSWYBInew · submitted 2026-06-04 · 💻 cs.AI

Amortizing Federated Adaptation: Hypernetwork Driven LoRA for Personalized Foundation Models

Sunny Gupta , Shambhavi Shanker , Amit Sethi This is my paper

Pith reviewed 2026-06-28 01:24 UTC · model grok-4.3

classification 💻 cs.AI
keywords federated learningLoRAhypernetworkspersonalizationfoundation modelsaggregation biasamortized adaptationnon-IID data
0
0 comments X

The pith

HyperLoRA replaces iterative per-client optimization and factor-wise averaging in federated LoRA with hypernetwork-generated initializations and product-space aggregation to remove structural bias and speed convergence.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces HyperLoRA as a framework for federated fine-tuning of foundation models that uses Low-Rank Adaptation. Current federated LoRA approaches incur structural aggregation bias because independently averaging the low-rank factors does not match the true combined update, and they suffer initialization lag because clients must reinitialize parameters each round. HyperLoRA instead trains a hypernetwork that maps client distribution signatures directly to LoRA initializations, amortizing the adaptation step. On the server a separate learned module synthesizes the aggregated update inside the low-rank product space rather than averaging factors, and a residual correction term stabilizes training under non-IID conditions. The result is reported faster convergence, reduced bias, and stronger personalization on vision and vision-language federated benchmarks.

Core claim

HyperLoRA employs a learned generator that maps client distribution signatures to LoRA initializations, amortizing per-client adaptation, and introduces a learned aggregation module that directly synthesizes updates in the low-rank product space, eliminating inconsistencies of factor-wise averaging, with a lightweight residual correction for non-IID distributions.

What carries the argument

Hypernetwork that generates client-specific LoRA initializations from distribution signatures together with a learned product-space aggregation module that operates on the combined low-rank update.

If this is right

  • Per-client adaptation cost drops because the hypernetwork produces usable initializations without iterative optimization on each client.
  • Aggregation bias disappears when updates are synthesized directly in the low-rank product space instead of averaged factor by factor.
  • Convergence accelerates because clients no longer reinitialize LoRA parameters at the start of every communication round.
  • Stability improves on heterogeneous data through the addition of a lightweight residual correction term.
  • Personalization performance rises on both vision and vision-language federated tasks relative to earlier LoRA federated baselines.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same hypernetwork-plus-product-space pattern could be applied to other parameter-efficient adapters beyond LoRA in distributed settings.
  • Learned aggregation operators may transfer to non-federated distributed training where low-rank updates are combined across nodes.
  • Because initialization is amortized, the approach could reduce total communication volume when scaling to very large foundation models.
  • The residual correction module suggests a route for combining learned operators with existing techniques for handling client drift.

Load-bearing premise

A hypernetwork can map client distribution signatures to effective LoRA initializations and a learned aggregation module can synthesize updates in the low-rank product space that eliminate the structural inconsistencies of factor-wise averaging.

What would settle it

On the same federated vision benchmarks, replace the product-space aggregation module with standard factor-wise averaging while keeping the hypernetwork generator, then measure whether convergence speed and personalization metrics fall back to the levels of prior federated LoRA methods.

Figures

Figures reproduced from arXiv: 2606.06154 by Amit Sethi, Shambhavi Shanker, Sunny Gupta.

Figure 1
Figure 1. Figure 1: HyperLoRA. Each client extracts a low-dimensional distribution signature si from its private dataset and feeds it to a shared hypernetwork generator Gϕ, which produces a personalized LoRA initialization (A0 i , B0 i ). After E steps of local optimization, the resulting low-rank factors (Ai, Bi) are uploaded. Rather than averaging factors independently which incurs structural bias (§3.2) the server applies … view at source ↗
read the original abstract

Federated fine-tuning of foundation models using Low-Rank Adaptation (LoRA) offers a communication efficient solution for distributed learning. However, existing federated LoRA methods suffer from two fundamental limitations: (1) structural aggregation bias, where independently averaging low rank factors fails to approximate the true combined update, and (2) client side initialization lag, as clients repeatedly reinitialize LoRA parameters across communication rounds, slowing convergence. We propose HyperLoRA, a unified framework that addresses both issues through amortized federated adaptation through hypernetwork-driven LoRA generation and product space aggregation. Instead of iterative per-client optimization, HyperLoRA employs a learned generator that maps client distribution signatures to LoRA initializations, effectively amortizing per client adaptation. On the server side, we introduce a learned aggregation module that directly synthesizes updates in the low-rank product space, eliminating the inconsistencies of factor-wise averaging. A lightweight residual correction module further improves stability under heterogenous (non-IID) client distributions.By replacing iterative optimization and heuristic averaging with learned operators, HyperLoRA jointly enables efficient personalization, unbiased aggregation, and faster convergence. Experiments on federated vision and vision-language benchmarks show that HyperLoRA achieves improved convergence speed, greater robustness to distribution shift, and stronger personalization performance compared to prior federated LoRA methods.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes HyperLoRA, a unified framework for federated fine-tuning of foundation models via LoRA. It targets two limitations of existing methods—structural aggregation bias from independent averaging of low-rank factors and client-side initialization lag—by employing a hypernetwork that maps client distribution signatures to LoRA initializations (amortizing per-client adaptation), a learned aggregation module that synthesizes updates directly in the low-rank product space, and a lightweight residual correction module for non-IID stability. The approach claims to replace iterative optimization and heuristic averaging with learned operators, yielding efficient personalization, unbiased aggregation, and faster convergence, supported by experiments on federated vision and vision-language benchmarks.

Significance. If the empirical claims hold and the learned operators deliver the stated benefits, the work would advance federated adaptation of foundation models by substituting heuristics with amortized, learned components, potentially improving communication efficiency, convergence, and personalization under heterogeneity. No machine-checked proofs, reproducible code, or parameter-free derivations are described.

major comments (2)
  1. [Abstract (framework description)] Abstract (framework description): the assertion that the learned aggregation module 'directly synthesizes updates in the low-rank product space, eliminating the inconsistencies of factor-wise averaging' is load-bearing for the unbiased-aggregation claim yet supplies no error bound, proof that the module output is closer to mean(BA) than avg(B)avg(A), or characterization of the function class needed to learn the product operation.
  2. [Abstract] Abstract: no equations, derivations, quantitative results, or experimental details (e.g., datasets, baselines, metrics, or tables) are provided, preventing verification of the claimed gains in convergence speed, robustness to distribution shift, and personalization performance.
minor comments (1)
  1. [Abstract] Abstract: the phrase 'amortizing federated adaptation' is used without a formal definition or citation to prior amortization literature.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive feedback. We respond to each major comment below and note planned revisions to the manuscript where appropriate.

read point-by-point responses

  1. Referee: [Abstract (framework description)] Abstract (framework description): the assertion that the learned aggregation module 'directly synthesizes updates in the low-rank product space, eliminating the inconsistencies of factor-wise averaging' is load-bearing for the unbiased-aggregation claim yet supplies no error bound, proof that the module output is closer to mean(BA) than avg(B)avg(A), or characterization of the function class needed to learn the product operation.

    Authors: We acknowledge that the manuscript provides no theoretical error bounds, proofs, or function-class characterization for the learned aggregation module. The approach is presented as an empirical alternative to factor-wise averaging, with the full paper demonstrating performance gains on federated benchmarks. We will revise the abstract to qualify the claim, indicating that the module mitigates inconsistencies via learned synthesis in the product space, as validated empirically rather than through formal guarantees. revision: yes

  2. Referee: [Abstract] Abstract: no equations, derivations, quantitative results, or experimental details (e.g., datasets, baselines, metrics, or tables) are provided, preventing verification of the claimed gains in convergence speed, robustness to distribution shift, and personalization performance.

    Authors: Abstracts are intentionally concise and omit detailed equations or results. The full manuscript contains the requested elements, including derivations of the hypernetwork and aggregation modules, experimental protocols, datasets, baselines, metrics, and quantitative tables in the methods and experiments sections. We will revise the abstract to include a brief statement summarizing the observed improvements in convergence and personalization on the reported benchmarks. revision: yes

standing simulated objections not resolved
  • Theoretical error bound or proof that the learned aggregation module output is closer to mean(BA) than factor-wise averaging, or characterization of the required function class.

Circularity Check

0 steps flagged

No derivations or equations present; claims rest on empirical validation without self-referential structure

full rationale

The abstract and framework description contain no equations, derivations, or self-citations. Central claims about hypernetwork-driven initialization and product-space aggregation are presented as design choices whose effectiveness is asserted via experiments, not derived from prior results or fitted parameters renamed as predictions. No self-definitional mappings, uniqueness theorems, or ansatzes smuggled via citation appear. The skeptic concern addresses missing formal bounds, which is a correctness issue rather than circularity. The derivation chain is self-contained against external benchmarks in the sense that no internal reduction to inputs is exhibited.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no free parameters, axioms, or invented entities are specified in sufficient detail to populate the ledger.

pith-pipeline@v0.9.1-grok · 5765 in / 1039 out tokens · 27226 ms · 2026-06-28T01:24:26.366000+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

84 extracted references · 13 canonical work pages · 4 internal anchors

  1. [1]

    Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS) , pages =

    Communication-Efficient Learning of Deep Networks from Decentralized Data , author =. Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS) , pages =

  2. [2]

    Proceedings of Machine Learning and Systems (MLSys) , year =

    Federated Optimization in Heterogeneous Networks , author =. Proceedings of Machine Learning and Systems (MLSys) , year =

  3. [3]

    International Conference on Machine Learning (ICML) , year =

    Ditto: Fair and Robust Federated Learning Through Personalization , author =. International Conference on Machine Learning (ICML) , year =

  4. [4]

    Personalized Cross-Silo Federated Learning on Non-

    Huang, Yutao and Chu, Lingyang and Zhou, Zirui and Wang, Lanjun and Liu, Jiangchuan and Pei, Jian and Zhang, Yong , booktitle =. Personalized Cross-Silo Federated Learning on Non-

  5. [5]

    Li, Xiaoxiao and Jiang, Meirui and Zhang, Xiaofei and Kamp, Michael and Dou, Qi , booktitle =. Fed

  6. [6]

    Karimireddy, Sai Praneeth and Kale, Satyen and Mohri, Mehryar and Reddi, Sashank and Stich, Sebastian and Suresh, Ananda Theertha , booktitle =

  7. [7]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Tackling the Objective Inconsistency Problem in Heterogeneous Federated Optimization , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  8. [8]

    International Conference on Learning Representations (ICLR) , year =

    Federated Learning Based on Dynamic Regularization , author =. International Conference on Learning Representations (ICLR) , year =

  9. [9]

    Federated Learning with Non-

    Zhao, Yue and Li, Meng and Lai, Liangzhen and Suda, Naveen and Civin, Damon and Chandra, Vikas , journal =. Federated Learning with Non-

  10. [10]

    Federated Learning on Non-

    Li, Qinbin and Diao, Yiqun and Chen, Quan and He, Bingsheng , booktitle =. Federated Learning on Non-

  11. [11]

    Foundations and Trends in Machine Learning , volume =

    Advances and Open Problems in Federated Learning , author =. Foundations and Trends in Machine Learning , volume =

  12. [12]

    Hu, Edward J and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , booktitle =. Lo

  13. [13]

    Zhang, Qingru and Chen, Minshuo and Bukharin, Alexander and He, Pengcheng and Cheng, Yu and Chen, Weizhu and Zhao, Tuo , booktitle =. Ada

  14. [14]

    Liu, Shih-Yang and Wang, Chien-Yi and Yin, Hongxu and Molchanov, Pavlo and Wang, Yu-Chiang Frank and Cheng, Kwang-Ting and Chen, Min-Hung , booktitle =. Do

  15. [15]

    , booktitle =

    Kopiczko, Dawid Jan and Blankevoort, Tijmen and Asano, Yuki M. , booktitle =. Ve

  16. [16]

    Meng, Fanxu and Wang, Zhaohui and Zhang, Muhan , journal =. Pi

  17. [17]

    Tian, Chunlin and Shi, Zhan and Guo, Zhijiang and Li, Li and Xu, Cheng-Zhong , journal =. Hydra

  18. [18]

    Parameter-Efficient Transfer Learning for

    Houlsby, Neil and Giurgiu, Andrei and Jastrzebski, Stanislaw and Morrone, Bruna and De Laroussilhe, Quentin and Gesmundo, Andrea and Attariyan, Mona and Gelly, Sylvain , booktitle =. Parameter-Efficient Transfer Learning for

  19. [19]

    Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP) , year =

    The Power of Scale for Parameter-Efficient Prompt Tuning , author =. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP) , year =

    2021

  20. [20]

    Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics (ACL) , year =

    Prefix-Tuning: Optimizing Continuous Prompts for Generation , author =. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics (ACL) , year =

  21. [21]

    Ben Zaken, Elad and Goldberg, Yoav and Ravfogel, Shauli , booktitle =. Bit

  22. [22]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learning , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  23. [23]

    European Conference on Computer Vision (ECCV) , year =

    Visual Prompt Tuning , author =. European Conference on Computer Vision (ECCV) , year =

  24. [24]

    International Journal of Computer Vision (IJCV) , volume =

    Learning to Prompt for Vision-Language Models , author =. International Journal of Computer Vision (IJCV) , volume =

  25. [25]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Conditional Prompt Learning for Vision-Language Models , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  26. [26]

    Khattak, Muhammad Uzair and Rasheed, Hanoona and Maaz, Muhammad and Khan, Salman and Khan, Fahad Shahbaz , booktitle =. Ma

  27. [27]

    Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey

    Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey , author =. arXiv preprint arXiv:2403.14608 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  28. [28]

    Nature Machine Intelligence , volume =

    Parameter-Efficient Fine-Tuning of Large-Scale Pre-trained Language Models , author =. Nature Machine Intelligence , volume =

  29. [29]

    Towards Building the Federated

    Zhang, Jianyi and Vahidian, Saeed and Kuo, Martin and Li, Chunyuan and Zhang, Ruiyi and Yu, Tong and Wang, Guoyin and Chen, Yiran , booktitle =. Towards Building the Federated

  30. [30]

    arXiv preprint arXiv:2403.12313 , year =

    Improving LoRA in Privacy-Preserving Federated Learning , author =. arXiv preprint arXiv:2403.12313 , year =

    work page arXiv

  31. [31]

    arXiv preprint arXiv:2409.05976 , year =

    Flora: Federated Fine-Tuning Large Language Models with Heterogeneous Low-Rank Adaptations , author =. arXiv preprint arXiv:2409.05976 , year =

    work page arXiv

  32. [32]

    arXiv preprint arXiv:2402.11505 , year =

    Federated Fine-Tuning of Large Language Models under Heterogeneous Language Tasks and Client Resources , author =. arXiv preprint arXiv:2402.11505 , year =

    work page arXiv

  33. [33]

    Bian, Jieming and Wang, Lei and Zhang, Letian and Xu, Jie , booktitle =. Lo

  34. [34]

    International Workshop on Federated Learning in the Age of Foundation Models, NeurIPS , year =

    Heterogeneous LoRA for Federated Fine-Tuning of On-Device Foundation Models , author =. International Workshop on Federated Learning in the Age of Foundation Models, NeurIPS , year =

  35. [35]

    Zhang, Zhuo and Yang, Yuanhang and Dai, Yong and Wang, Qifan and Yu, Yue and Qu, Lizhen and Xu, Zenglin , booktitle =. Fed

  36. [36]

    Kuang, Weirui and Qian, Bingchen and Li, Zitao and Chen, Daoyuan and Gao, Dawei and Pan, Xuchen and Xie, Yuexiang and Li, Yaliang and Ding, Bolin and Zhou, Jingren , booktitle =

  37. [37]

    arXiv preprint arXiv:2308.06522 , year =

    SLoRA: Federated Parameter Efficient Fine-Tuning of Language Models , author =. arXiv preprint arXiv:2308.06522 , year =

    work page arXiv

  38. [38]

    Yi, Liping and Yu, Han and Wang, Gang and Liu, Xiaoguang and Li, Xiaoxiao , journal =

  39. [39]

    Wu, Feijie and Li, Zitao and Li, Yaliang and Ding, Bolin and Gao, Jing , booktitle =. Fed

  40. [40]

    and Le, Quoc V

    Ha, David and Dai, Andrew M. and Le, Quoc V. , booktitle =. Hyper

  41. [41]

    International Conference on Learning Representations (ICLR) , year =

    Continual Learning with Hypernetworks , author =. International Conference on Learning Representations (ICLR) , year =

  42. [42]

    4th Workshop on Meta-Learning, NeurIPS , year =

    Meta-Learning via Hypernetworks , author =. 4th Workshop on Meta-Learning, NeurIPS , year =

  43. [43]

    International Conference on Machine Learning (ICML) , year =

    Personalized Federated Learning Using Hypernetworks , author =. International Conference on Machine Learning (ICML) , year =

  44. [44]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Layer-Wised Model Aggregation for Personalized Federated Learning , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  45. [45]

    Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics (ACL) , year =

    Parameter-Efficient Multi-Task Fine-Tuning for Transformers via Shared Hypernetworks , author =. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics (ACL) , year =

  46. [46]

    , booktitle =

    He, Yun and Zheng, Steven and Tay, Yi and Gupta, Jai and Du, Yu and Aribandi, Vamsi and Zhao, Zhe and Li, Yaguang and Chen, Zhao and Metzler, Donald and Cheng, Heng-Tze and Chi, Ed H. , booktitle =. Hyper

  47. [47]

    , booktitle =

    Ivison, Hamish and Peters, Matthew E. , booktitle =. Hyperdecoders: Instance-Specific Decoders for Multi-Task

  48. [48]

    , booktitle =

    Ivison, Hamish and Bhagia, Akshita and Wang, Yizhong and Hajishirzi, Hannaneh and Peters, Matthew E. , booktitle =

  49. [49]

    International Conference on Machine Learning (ICML) , year =

    Hypertuning: Toward Adapting Large Language Models without Backpropagation , author =. International Conference on Machine Learning (ICML) , year =

  50. [50]

    Lv, Chuanyang and Li, Liang and Zhang, Shuai and Chen, Gehui and Qi, Fanchao and Zhang, Ningyu and Zheng, Hai-Tao , booktitle =. Hyper

  51. [51]

    , booktitle =

    Charakorn, Rujikorn and Cetin, Edoardo and Tang, Yujin and Lange, Robert T. , booktitle =. Text-to-Lo

  52. [52]

    , journal =

    Charakorn, Rujikorn and Cetin, Edoardo and Uesaka, Shinnosuke and Lange, Robert T. , journal =. Doc-to-Lo

  53. [53]

    International Conference on Learning Representations (ICLR) , year =

    Generative Adapter: Contextualizing Language Models in Parameters with a Single Forward Pass , author =. International Conference on Learning Representations (ICLR) , year =

  54. [54]

    arXiv preprint arXiv:2506.06266 , year =

    Cartridges: Lightweight and General-Purpose Long Context Representations via Self-Study , author =. arXiv preprint arXiv:2506.06266 , year =

    work page arXiv

  55. [55]

    Li, Yichuan and Ma, Xiyao and Lu, Sichao and Lee, Kyumin and Liu, Xiaohu and Guo, Chenlei , booktitle =

  56. [56]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Learning to Compress Prompts with Gist Tokens , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  57. [57]

    2016 , eprint=

    HyperNetworks , author=. 2016 , eprint=

    2016

  58. [58]

    International Conference on Machine Learning (ICML) , year =

    Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks , author =. International Conference on Machine Learning (ICML) , year =

  59. [59]

    On First-Order Meta-Learning Algorithms

    On First-Order Meta-Learning Algorithms , author =. arXiv preprint arXiv:1803.02999 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  60. [60]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Fast and Flexible Multi-Task Classification Using Conditional Neural Adaptive Processes , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  61. [61]

    International Conference on Machine Learning (ICML) , year =

    Perceiver: General Perception with Iterative Attention , author =. International Conference on Machine Learning (ICML) , year =

  62. [62]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Personalized Federated Learning with Moreau Envelopes , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  63. [63]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Personalized Federated Learning with Theoretical Guarantees: A Model-Agnostic Meta-Learning Approach , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  64. [64]

    International Conference on Machine Learning (ICML) , year =

    Exploiting Shared Representations for Personalized Federated Learning , author =. International Conference on Machine Learning (ICML) , year =

  65. [65]

    Oh, Jaehoon and Kim, SangMook and Yun, Se-Young , booktitle =. Fed

  66. [66]

    Federated Learning with Personalization Layers

    Federated Learning with Personalization Layers , author =. arXiv preprint arXiv:1912.00818 , year =

    work page internal anchor Pith review Pith/arXiv arXiv 1912

  67. [67]

    IEEE Transactions on Neural Networks and Learning Systems , year =

    Towards Personalized Federated Learning , author =. IEEE Transactions on Neural Networks and Learning Systems , year =

  68. [68]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Federated Learning from Pre-trained Models: A Contrastive Learning Approach , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  69. [69]

    Tian, Yuanyishu and Wan, Yao and Lyu, Lingjuan and Yao, Dezhong and Jin, Hai and Sun, Lichao , journal =. Fed

  70. [70]

    arXiv preprint arXiv:2210.01708 , year =

    Conquering the Communication Constraints to Enable Large Pre-Trained Models in Federated Learning , author =. arXiv preprint arXiv:2210.01708 , year =

    work page arXiv

  71. [71]

    arXiv preprint arXiv:2402.06954 , year =

    Federated Learning of Foundation Models: A Survey , author =. arXiv preprint arXiv:2402.06954 , year =

    work page arXiv

  72. [72]

    arXiv preprint arXiv:2305.11414 , year =

    Federated Foundation Models: Privacy-Preserving and Collaborative Learning for Large Models , author =. arXiv preprint arXiv:2305.11414 , year =

    work page arXiv

  73. [73]

    Transactions on Machine Learning Research (TMLR) , year =

    Personalization of Large Language Models: A Survey , author =. Transactions on Machine Learning Research (TMLR) , year =

  74. [74]

    A General Language Assistant as a Laboratory for Alignment

    A General Language Assistant as a Laboratory for Alignment , author =. arXiv preprint arXiv:2112.00861 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  75. [75]

    Propagating Knowledge Updates to

    Padmanabhan, Shankar and Onoe, Yasumasa and Zhang, Michael and Durrett, Greg and Choi, Eunsol , booktitle =. Propagating Knowledge Updates to

  76. [76]

    arXiv preprint arXiv:2209.15189 , year =

    Learning by Distilling Context , author =. arXiv preprint arXiv:2209.15189 , year =

    work page arXiv

  77. [77]

    Structure and Interpretation of Computer Programs

    Harold Abelson and Gerald Jay Sussman and Julie Sussman. Structure and Interpretation of Computer Programs. 1985

    1985

  78. [78]

    Visual Information Extraction with Lixto

    Robert Baumgartner and Georg Gottlob and Sergio Flesca. Visual Information Extraction with Lixto. Proceedings of the 27th International Conference on Very Large Databases. 2001

    2001

  79. [79]

    Brachman and James G

    Ronald J. Brachman and James G. Schmolze. An overview of the KL-ONE knowledge representation system. Cognitive Science. 1985

    1985

  80. [80]

    Complexity results for nonmonotonic logics

    Georg Gottlob. Complexity results for nonmonotonic logics. Journal of Logic and Computation. 1992

    1992

Showing first 80 references.